metadata
language: en
library_name: pytorch-image-translation-models
pipeline_tag: image-to-image
tags:
- image-to-image
- diffusion
- DiffuseIT
- FFHQ
- face
- identity-preservation
- text-guided
DiffuseIT: FFHQ 256
Face-focused diffusion model with identity preservation. Pre-trained on FFHQ 256×256.
Source: cyclomon/DiffuseIT — converted from ffhq_10m.pt
Model Description
- Architecture: Guided diffusion (OpenAI-style UNet, face-optimized)
- Resolution: 256×256
- Task: Face image translation with identity preservation (use
use_ffhq=True) - Self-contained: Includes
id_model/(ArcFace IR-SE50) for identity loss
Usage
from examples.community.diffuseit import load_diffuseit_community_pipeline
pipe = load_diffuseit_community_pipeline(
"BiliSakura/DiffuseIT-ckpt/ffhq-256",
use_ffhq=True,
diffuseit_src_path="projects/DiffuseIT",
)
pipe.to("cuda")
out = pipe(
source_image=face_img,
prompt="Target description",
source="Source description",
use_range_restart=True,
output_type="pil",
)
Citation
@inproceedings{kwon2023diffuseit,
title={Diffusion-based Image Translation using Disentangled Style and Content Representation},
author={Kwon, Gihyun and Ye, Jong Chul},
booktitle={ICLR},
year={2023},
url={https://arxiv.org/abs/2209.15264}
}